The improvement of performance among athletes and training techniques has become integral focus in sports analytics optimizing outcomes and maximizing potential. On the contrary, the traditional approach lacks data dependency and struggles to provide actionable recommendations. This work proposes a novel approach to apply reinforcement learning (RL) algorithms for optimizing athlete performance through sports analytics datasets to make decisions of training strategy. Initially, sports data such as game results and player statistics are collected for the development of analytical models to interpret and utilize player data. This study specifically examines athlete performance using the Beginners Sports Analytics NFL Dataset, which examines player tracking and game events from American football. After that, the data undergoes cleaning through Singular Value Decomposition (SVD) for dimensionality reduction and Local Outlier Factor (LOF) for outlier detection. After this, the cleaned data is then used to extract feature via Dynamic Time Warping capable of aligning the sequences of player performance over time. Finally, a Deep Q-learning algorithm integrated with Neural Architecture Search (DQ-NAS) is utilized for accurate decision-making toward optimizing athlete performance and training strategies. The agent is rewarded based on performance, The performance is measured by improvements in key athlete metrics such as speed, endurance, skill execution, and reduction in errors (e.g., mistakes made during play). The results, showing a 7–15% increase in speed, endurance, and skill, also 10–20% reduction in mistakes. The rewards lead to a 15% increase in performance over time using the DQ-NAS model. Across experiments, framework achieved a 7–15% improvement in training skill efficiency and a 15% cumulative reward gain over baseline methods. Nevertheless, the conclusions should be interpreted carefully because the Beginners Sports Analytics Dataset is limited, contains no real-world athlete variability, and may not sufficiently capture the training context for an elite sports firm.
Yuchen Gui (Sat,) studied this question.